\begin{table}[H] \centering
\newcolumntype{R}{>{\raggedleft\arraybackslash}X}
\newcolumntype{L}{>{\raggedright\arraybackslash}X}
\newcolumntype{C}{>{\centering\arraybackslash}X}

\caption{Policy leverage}
\label{tab:het_policy_leverage}
\begin{tabularx}{\linewidth}{@{}lCCCCCC@{}}

\toprule
 & \multicolumn{2}{c}{\textbf{(Minus) Inflation}} & \multicolumn{2}{c}{\textbf{(Minus) Unemployment}} & \multicolumn{2}{c}{\textbf{(Minus) Inflation}} \tabularnewline & (1) & (2) & (3) & (4) & (5) & (6)  \tabularnewline 
{}&{High inflation}&{Low inflation}&{High emp.}&{Low emp.}&{High CBI}&{Low CBI} \tabularnewline
\midrule \addlinespace[\belowrulesep]
Electoral turnover&0.92***&-0.06&0.71**&0.04&0.44&1.05*** \tabularnewline
&(0.39)&(0.07)&(0.34)&(0.17)&(0.29)&(0.47) \tabularnewline
\midrule N&988&899&681&650&663&600 \tabularnewline
\bottomrule \addlinespace[\belowrulesep]

\end{tabularx}
\\ \parbox{\linewidth}{\footnotesize \caption*{\footnotesize \emph{Notes}: In column (1) (resp., 2), we report the estimated effect of electoral turnovers on (minus) inflation when inflation on the year before the election is above (resp., below) the median. In column (3) (resp., 4), we report the estimated effect of electoral turnovers on (minus) unemployment when unemployment on the year before the election is above (resp., below) the median. Finally, in column (5) (resp., 6), we report the estimated effect of electoral turnovers on (minus) inflation when the independence of the central bank \citep[measured by][]{garriga2016central} on the year before the election is above (resp., below) the median. In each case, we measure the median in the subsample of elections with a running variable between -15pp and +15pp. Using the method of \cite{clogg1995statistical}, we can reject the equality of the estimates of columns (1) and (2) (p-val. =  0.013), of the estimates of columns (3) and (4) (p-val. =  0.076), but not of the estimates of columns (5) and (6) (p-val. =  0.267). We obtain broadly consistent results when running a parametric regression in which we include the interaction between the treatment and the dimension of heterogeneity.}}
\end{table}
